Personal injury case value calculator.

How the valuation model works

Predict runs your inputs through a gradient-boosted regression trained on over 20,000 precedent cases and 200,000 data points from 2018 forward. It returns a confidence-banded gross case value — a point estimate plus the 90% interval, before attorney fees, case costs, and medical liens — alongside the comparable-verdict cohort and the jurisdiction tuning that produced the number.

The model is plaintiff-side only. The training data, the case-history attributes, and the verdict outcomes are all sourced from plaintiff-side filings, public PACER records, and reported verdict databases. There is no claims-side carrier data in the model, and there never will be. Plaintiff-side only is the trust contract; the dataset reflects it.

The output works across the case lifecycle. 60 seconds, not two hours. Intake and case selection is one moment — type the case facts you'd type into a Verdict Search lookup and get a gross case value with a band. The same valuation re-runs as the file matures: with the police report and medical records in hand for demand-letter prep, and again heading into pre-litigation negotiation and settlement.

Inputs that actually move the number

Six fields a plaintiff attorney supplies at intake. Injury severity and direct damages do the most work, with jurisdiction close behind. The model handles dozens of secondary signals — county demographics, plaintiff age and gender, commercial-vehicle involvement, per-county verdict medians — internally.

• Injury severity. Minor, moderate, severe, catastrophic — the single largest factor. Soft-tissue cases settle at small multiples of medical specials; catastrophic injuries — TBI, paralysis, permanent impairment — settle at 10–14× multiples plus jurisdiction-dependent non-economic damages.
• Jurisdiction. A major swing across states — second only to the injury itself. The same fact pattern resolves to a different number in different states, because local verdict history and county-level jury behavior diverge. The model trains a separate jurisdiction fold per state and, where data density allows, per county, so each prediction is tuned to where the case will actually be tried.
• Medical specials. The anchor for the damages calculation. The classic "multiple of meds" heuristic is a reasonable first approximation, but the multiplier itself varies by severity and jurisdiction — which is why a static spreadsheet misses by 30–50% on cases the model lands inside the band.
• Case context. Predict returns a gross value — before any comparative-fault reduction — so a contested case and an admitted one score the same at the same injury and jurisdiction. The comparative-fault haircut is yours to apply, not the model's.
• Property damage. A weak signal on case value, but a useful one for sanity-checking severity claims. The model uses it as a correction term, not as a primary driver.
• Case type. MVA versus premises liability. The same medical-specials profile produces materially different case values across case types — premises cohorts resolve at lower multiples than MVA.

The calculator on this page accepts all six. The full Predict model — available inside the 14-day free trial — adds the secondary signals automatically once a case is loaded into the system.

Why every number ships with a confidence band

The instinct, when building a pricing model for attorneys, is to suppress uncertainty. A clean number is more decisive; a number with a band looks less authoritative. We almost did that. We were talked out of it by attorneys.

Attorneys are trained to evaluate uncertainty. It's the job. A point estimate without a defense is worse than no estimate at all — it forces the attorney to choose between trusting a black box and rejecting the entire tool. A confidence band is the defense. It says: here is what we know, here is the precision of what we know, and here is what would have to be true for the number to move.

The number is the headline. The band is the methodology. Never lead with the range.

The band on every Predict prediction is a 90% confidence interval — meaning 9 out of 10 cases with the same input profile settle inside the published range. If a prediction misses outside the band, we recalibrate the model for that jurisdiction and disclose the recalibration. That's a brand commitment, not a marketing line.

By case type — MVA and premises liability

The two case types Predict is calibrated for cover the bulk of plaintiff PI volume. Each has its own modeling considerations:

• Motor vehicle accidents (MVA). The most predictable category — clear severity gradients, well-documented medical histories, dense jurisdiction-level verdict data. Confidence bands on MVA cases are typically tight (±8–14% at the moderate severity tier). The MVA case value calculator walks through the case-specific factors and surfaces typical comp ranges.
• Premises liability (PL). Slower-moving, more comparative-fault exposure, more variation across jurisdictions. Confidence bands run modestly wider (±12–18% at moderate severity). The premises liability case value calculator handles slip/trip and fall, inadequate security, and building-maintenance cases.

The calculator on this page handles both case types. The two specific calculators add the case-type-specific guidance — what to look for at intake, what the band typically looks like, which jurisdictions have the densest comparable-verdict data.

Why jurisdiction is the biggest cross-state factor

The injury itself is the largest driver of value. But holding the injury constant, state-level PI economics still diverge widely. The same fact pattern — a moderate cervical strain from a low-speed rear-end — resolves to a different number in different states, because local verdict history and county-level jury behavior diverge. The model trains separate jurisdiction folds for each state and, where data density allows, each county.

For a per-state view of jury-verdict density and how each jurisdiction is tuned, see the state-by-state calculator hub. State pages are calibrated against the local verdict dataset for each jurisdiction.

Where Predict fits alongside Verdict Search, evaluator memos, and gut

Predict is the valuation spine on every case. The other tools sit around it: a verdict reporter is a citation source you pull named-case text from, and a paid evaluator memo is an optional second opinion on a rare, genuinely novel outlier. Here is how they relate at the decision moment:

• Doing nothing — gut + spreadsheet. The largest competitor. The status quo for most solo and small-firm PI attorneys. Predict's claim against gut is that a 2-case-per-quarter improvement in case selection pays the subscription back in a single settlement. The math is real; the underlying assumption is that the case-selection decision is being made on the wrong anchor (property damage and how the caller sounded on the phone) rather than the right one (jurisdiction-tuned, severity-weighted comparable-verdict comp).
• Verdict Search / Jury Verdict Reporter. Industry-standard lookup databases — a citation source, not a model. You pull named-case text from them to cite the comparables behind a number. Predict does the weighting first and returns a jurisdiction-tuned value with its band; the reporter is where you go to quote the specific cases underneath it.
• Outsourced settlement-evaluator firms. Paralegal-staffed firms that produce a per-case demand-letter memo. Useful as an optional second opinion on a rare, genuinely novel outlier — but slow (one to two weeks) and expensive (~$1,500 per case), so it doesn't scale to the everyday file. Predict is the valuation spine that runs in 60 seconds on every case, inside the workflow, on a $499 per-seat / month subscription.

Frequently asked questions